When measuring capacitor values using the RC like this, it's handy (on the analog scopes) to adjust the variable vertical scale so that the full amplitude covers exactly 8 divisions. Then, one RC will be when the voltage crosses the 5th division. Easy.
Something that needs to be mentioned is that as you increase the physical size of capacitor (and increase the voltage rating, and use a good dialectric like X7R) the dc bias become less of an issue. This is very helpful when doing parametric searches. For instance, suppose you need about 10uf at 5volts. A 10v 805 (Samsung CL21B106KPQNFNE) will lose about 50% of its capacitance but a 25v 1206 (CL31B106KAHNNNE) will only lose 20%. Both caps cost about the same and have good graphs on the characteristics sheet on digikey.
The recent explosion of high value small ceramic capacitors has brought this issue to the forefront. The phenomenon is due to polarization of the dielectric, and it tends to get worse as you increase capacitance and voltage rating and decrease the size of the cap. Typically as another commenter said, if you have two caps with the same capacitance and voltage rating, the smaller one will have a bigger DC bias dependency. Part of the reason is that in order to get high capacitance and voltage rating in a small package, they switch to dielectrics like barium titanate that, while withstanding higher voltages over a smaller distance, are much easier to polarize.
A phenomenon I've noticed over 25 plus years of servicing but I never knew why until now. I now know that it's not my imagination playing tricks on me. Thanks Dave! You've out did yourself...!
Reminds me of a circuit we built for a customer that used lots of RC delays that kept failing on test because the designer had not allowed for the bias voltage of the transistors that switched on the RC circuit. We also had trouble with the dielectric aging, e.g. x7r can change by 2% after 24 hours after its been soldered (and heated above the curie point)
That's indeed very much true. Apart from the short term phenomenon Dave demonstrated in MLCCs, there is als a long term decrease of capacitance, I measured it myself during a month and the capacitance does indeed keep decreasing. When you heat up the capacitor afterwards, it gets its original capacitance back. I read somewhere for X7R and X5R the loss is 2.5% per decade hour and for Y5V it is 7% per decade hour
Hugo Coolens Never a good idea to use x7r caps in timing circuits, but it is good fun showing the calculations to the designer on why his circuit will not meet the test limits he has specified.
I was a product development electronics engineer for almost 20 years and never knew this. A real eye opener. When actual capacitance values matter ... pay attention!
Thanks Dave! This video really helped me with some op-amp filter design problems. Thought it was flat-out sensitivity rather than DC bias voltage. You're awesome!!!! KEEP IT UP!!!!
Great video, thank you very much! And very useful: the 555 timer has voltage on C changing from 1/3 Vcc to 2/3 Vcc, so it's always with that bias, and never the full capacity.
Mac Carter: Even though Voltage is relative, when you have a capacitor with a bias voltage, the dielectric has an electric field applied to it by the bias voltage, not the difference as you would expect... This electric field causes the dielectric to expand, and in capacitors the farther the plates are from each other the less capacitance you get.
Truly of the best teachers, in depth knowledge and fine details being explained, beautiful lives and better world building depends on such beautiful teachers, May God bless 🙌 with more such teachers, God bless you Sir..
That is why you don't want to use ceramics in audio applications. They really add distortion. ;) You said electrolytics do not have the issue, but they do. Probably not as much as they used to though, I seem to recall most changes happen under 1V. I no longer see cap/volt curves in the data sheets. I remember old radios had electrolytics in their AGC circuits reverse-polarized. They got more capacitance out of it and there was something about the non-linear response. I had a client use some MCC ceramics in a headset amp as they had those in stock instead of the ones I specified. They made great microphones ;)
They can be used, but are quite expenive for the voltage/capacitance ratings (the drawback is the piezo-effect), the only thing to keep in mind is to use a higher capacitor voltage ratio, for exemple on some of my designs I use 50V capacitors on 5V filtering and they only have a tolerance deviation of just 10% and this covers the Voltage deviation+ temperature deviation , the total maximum deviation it should be 10%(initial)+10% (temp, Volt)....also ceramics have low esr..a 10V 10uF should have a few mOhms...
Old video, but one of your most relevant still! Unfortunately this video did not save me from making a split second decision for bom changes due to availability and ended up with my capacitance being reduced by 75% at my supply voltage. Needless to say, supply-filtering was not great.
I just faced with this issue yesternday Dave :) then You know what I used? I found some piece old (from the farao's age i think) hand soldered biffy big bug cap (i will send few pieces to You, interesting!) and bang on. no changes on nothing :) the circuit is used for detect mm size metal thinks in the wood, and the oscillator is working in a wood workshop. dust, AC kicks, temperature changing, everything. And today I changed the full oscillator to tube. bang on. with modern caps, even not detect nothing :) with old caps... well almost detect a bug inside the wood just with different reaction voice. go figur! off course in this case the size of the osc is not mater:)
I have struggled for the past couple of weeks trying to make a frequency circuit that is able to go from 50hz-20khz and I have had stability problems with every circuit. The whole time I was using the wrong caps. Thanks for your help, and turning me in the right direction.
greate episode - I could have used that informations two week agos. We have new component engineer at my place and I as an PCB Designer Engineer was forced to give an introduction into... well: Electronics. As I am not that deeply into components, I was the wrong pick but I guess the best option given the alternatives. Now the fun part next on the road side: My manager wants to use altium for circuit simulation. Well, don´t get me started with Intergrated circuits, but based on this information - all things ignored about Altium sucking at simulation - even capacitors can not be simulated with reasonable effort and/or accuracy.
Fascinating, I have been aware of some of the strange characteristics of various types of capacitor over the years but had totally overlooked this effect. I was aware of the piezo electric effects some capacitors exhibit, the self resonant effects seen in spiral wound aluminium electrolytics and some other strange memory effects.... The effect you demonstrate has serious repercussions if a coupling capacitor carrying dc bias is used to also provide LF roll-off. Thank you for uploading this.
7:08, when you explain the data sheet, the 1+/-0.1 kHz does not mean that the capacitance of the part should be measured at 100Hz, but with a frequency between 900 and 1100 Hz, with a Voltage of 1.0 Volts. Still, these huge variations, not just on the capacitances, but on the charging waveshape as well, are a huge surprise to me. Never expected anything like this!
There's nothing to stop this being in a textbook. For example, it's in The Art of Electronics (albeit in the X Chapters volume). The trouble is, you have to actually have to read the book... and who has time for that! (OK, I realize this comment is from 9 years ago hahaha.)
nice one, radio guys learn these quirks early on, but these days speeds of even PSUs switch speeds are rising so high that almost anyone can get caught out even on small caps in mundane circtuis, never mind in RF oscillators or timing kit. Another bugbear of mine is people not undertanding the workings of electrolytic caps. Aluminium ones, in particular, vary capacitance with applied voltage, so if undervolted compared to their working voltage rating can produce significantly lower capacitance than the marked value. I cringe when "badcaps" people say: "so I desoldered the old bad 100uF 16v and fitted a 100uF 63volt, because that'll last longer" then wonder why it fails after 2 month of huge ripple and bad hum :|
Part of the error is due to the voltage droop in the square wave due to charging current drawn. Add the output resistance of your function generator (which is also in series with your 1 k) to get a better time constant measurement. Typically, this is 50 Ohms so C=10.6 ms / 1.05kOhm gives a much more likely 10.095 uF. This is independent of the fact that ceramics' dielectric constant is electric field dependent. Also, V(t)=V_start + DV(1 - e^(-t/RC)).
Another issue people miss is the fact that if a voltage is applied to a cap , an equal and opposite voltage appears on the other terminal, this is for so called dc blocking cap applications. If it it is connected to a high impedance source this an issue. The op amp or circuit can still be damaged by a dc voltage causing a negative voltage spike.
Hi Dave :-) I think that the formula at 14:17 (with the annotation fix) is wrong. The general (and correct) formula for the voltage across the capacitor in this circuit is : Vc(t) = Vc(+inf) + [Vc(0+) - Vc(+inf)]*e^-(t/thau) Where Vc(+inf) is Vfinal and Vc(0+) is Vstart. This is derived from the differential equation of the RC circuit.
thanks Dave ! i was wondering why the modern 100nF caps (apparently class2 according to the dimensions) i use for decoupling have very similar decoupling effect like my old stock (15+ years) of 10nF ones ;) now i know the answer , so thanks again and keep going !
Holy crap! I just made RC filters for my project with all caps being ceramic instead of electrolytic and this video pops up. UA-cam starts to scare me! D:
You´re just pointed out why we do not use ceramic caps in any audio circuits path. As the insulator between the foils witch forms the cap is changing its values over bias voltage....
I think the term for this is "derating" - so the graph Dave showed of change in capacitance against voltage is a derating curve. Apologies if someone else already said this!
Great video, I knew there were some variances in these caps by looking at datasheets but it would be interesting to know more about the physics of why they change so drastically under these conditions. The materials used under DC bias conditions and various scenarios which lead to this variance would be a good video to get more in depth with the processes going on.
It is basically the electric dual of magnetic saturation. To make high-value capacitors physically small you use dielectrics with high relative permittivity (e.g. Barium Titanate). There is no free lunch though, the extra polarisation comes from physical reconfiguration of the charges in the dielectric material, and there is a limit to how much you can polarise a dielectric before it can help you no more (saturation - a state as ordered and contributing to the applied field as it can be in its current phase).
Great demonstration! This can really bite you if you are using caps to absorb transient voltages. We use ceramics right at connectors for EMI and ESD protection. I have been bit by this before in the past because i didnt compensate for this drop in capacitance and the caps kept exploding. Frown!
Was having this issue in the grid feed inverter I'm working on. I put some RC snubbers on the MOSFETs and could only find ceramics at hand, and it didn't improve the ringing as calculated (did reduce it, but it was still there). After ordering in some polypropylene film caps (Vishay MKP385), now no ringing! I suspected that it was this effect along with better temperature characteristics of poly caps perhaps.
Variable naming in the formula at 14:17 isn't right. if Vstart = 0, then you the multiply the entire expression by 0, and the formula says the voltage will always be 0 which clearly isn't right. For the step input, V0 isn't zero but the step voltage i.e.5 volts in this example; plot y = 5(1 - e^(-t/RC)) and that'll be clear. My guess for the DC offset is that it should be V = Voffset + Vstep(1 - e^(-t/RC)), i.e. 5 + 1(1 - e^(-t/1000C)) in this case...?
Small, cheap, stable - pick any two. The only truly linear dielectric is vacuum, and even that will eventually screw you with tunnelling at close spacing or vacuum breakdown and pair production at more extreme conditions.
Thanks for that Dave, but I wonder, could you do a video about how to choose the thickness of a wire? I am having some issues with it. (according to an online calculator it should'ave worked, but the wire ended melting the insulation, shorting over and destroying my brand new digital temp sensor...)
Very informative, thanks. Did you, by chance, decide to address this topic based on my recent comment on one of your older videos about caps where I warned about this problem?
EEVblog Never mind. The minds of people of a feather seem to be surprisingly aligned sometimes. Not that I'll ever be in the same league as you. Amazing job.
I think you misread the table at 7:13. It should have been the second row. Also, the cap should have been tested at 1khz+-0.1 So between 900 and 1100hz, whether it was the first or second row. The voltage should have been 0.5v+-0.1. I think so anyway! Sorry to be so nit-picky. But it's what I'm like.
Reminds me of how the gate charge on a MOSFET behaves... The gate appears as a capacitance, but that increases with the voltage, so whilst the gate may appear as ~1nF at a 0-5V signal, a 0-7V signal may see a ~3nF capacitance; this is usually written as "gate charge" measured in coulombs, and is typically given at 5V. -some- datasheets give an additional level, maybe 10V... The increase in gate charge is pretty drastic, and not many datasheets have it marked outside the standard 5V level, and i'm yet to find a graph that illustrates the gate charge Vs gate voltage characteristic. And as dave said, don't even mention the effects of frequency... That's just a whole dimension of changes on it's own :/
Hi Dave, warm regards from cold Russia! I tried to verify your info with my RLC meter by splitting voltage on 2 caps and it is true, the capacitance drops tremendously. If you have 5 minutes you can look at my video report named "Ceramic Capacitance" on my channel (can't add video reply though). Thank you for sharing knowledge with us!
Dave, I think you have a little typo in the vid description "If your 10uF capacitor really 10uF in your circuit?" but I think you meant Is. ;) Btw, great video as always! Keep up the great work :)
You can put that 6% down to the tolerance on the resistor too, no? I didn't see you measuring the 1k resistor's actual value, which definitely plays into the calculation of tau. Not that it makes a significant difference to the point you're making, of course.
As much as this is super interesting. This methodology is flawed. You cannot really use RC constant method, if the charging is not in the shape of 1-e^(-tau/t). What you can do instead, is take a log(1 - V/Vfinal) and fit a line a*t+b, and from average 'a' find out capacitance over range, or log(1 - V/Vfinal)*t, and see how capacitance "changes" as it MLCC is being charged. (i.e. it is operating in higher voltage region). Similar for discharge. In fact this will show variations in "capacitance" in single shot without need to change step response in signal generator. I am certain these effects can be pretty accurately modeles as parasitic inductances and capacitances, and they relate directly to physical reasons why we see these shapes on scopes.
Second link (avx) now a deadlink on avx site. The next avx link has similar problem. Wawawah That helps explain why my 3 means used to measure ceramic caps vary so much. Thanks!
in the 5V biased case, the voltage applied across the cap is the same 6V, but according to murata's QA here www.murata.com/support/faqs/products/capacitor/mlcc/char/0005 and a lot other threads they treat 'DC bias voltage' as voltage across capacitor, and capacitance decreases when voltage increases, well in Dave's case the voltage across the cap stays the same, so apparently they are not the same thing, or are they?
You didn't test the capacitor at voltages and frequency according the manufacturer's specifications. The data sheet at 7:15 specifies that test RMS voltage should be between 0.4 and 0.6 V RMS (line 2 column 2 for a 6.3 V cap). You tested the caps at 5 V square wave with duty cycle of 20% which corresponds to an RMS value of 2.24 V. Four times the recommended voltage. According to column 1 line 2 it should also have been tested between 900 and 1100 Hz. You tested it at 5 Hz? Now that's a greater challenge if you want to see the entire charge cycle on the scope but would be possible at 1 kHz with a 10 ohm resistor given a 10 microF cap.
Excellent Dave! Thank you!! And a bit of quibbling... in the case where you had a 0 to 10V, 20% duty cycle waveform, you do have a bias, right? because the average value is, what? 2 V ? (Don't know the exact value... it's been too many years...)
EEVblog Your 0-10V signal is biased. It has a substantial DC component. To be unbiased you would need to swing from negative to positive in a way that eliminates the DC component. Many electronics engineers are in total denial about the weird behaviour of capacitors, and accuse you of being an audiophile when you try to explain these effects to them. This is one of the very few areas where the audiophiles actually have it right. Capacitors are seriously funky, and it really matters which ones you use. Those DC related effects you are seeing are non-linear. Another funky polarisation effect with capacitances happens with Kapton flexible PCBs. The strength of the effect varies with the exact variety of Kapton, but I think it is always there. Measure the capacitance of some traces on a flexible PCB. Touch the PCB for a moment and the capacitance goes up. Let go and the capacitance falls to its original value.Now touch and keep touching the PCB for a couple of minutes, and then let go. The capacitance falls as you let go, but to a value distinctly higher than its original level. Over the next minute or so the capacitance gradually falls to its original level. This is not due to your finger warming the Kapton. You can warm the Kapton with hot air to around body temperature and you won't see the kind of capacitance change a sustained touch causes. The Kapton flexible PCBs which Apple use for cap-touch pads in the ipods only show this effect to a minor extent. I think they carefully selected the type of Kapton they use for this very characteristic. I have seen other products with Kapton cap-touch sensors where the effect is so strong that the adaptive capacitance tracking algorithms are only just about able to cope with the change.
Those tricky little bastards. I would never guess the cause if something like this would happen in real life circuit. On the other hand, I wonder if those caps could be used like varicap diodes.
Ok, so the capacitance changes with offset voltage but what is really changing? The physical dimensions and construction of the capacitor plates which calculate to 10 uf are not changing, yet the capacitance is. How can this be? Is the actual internal capacitor plates physically expanding/ shrinking to cause this effect?
When measuring capacitor values using the RC like this, it's handy (on the analog scopes) to adjust the variable vertical scale so that the full amplitude covers exactly 8 divisions. Then, one RC will be when the voltage crosses the 5th division. Easy.
Hey that's handy to know. 62.5 is pretty close. Thanks!
I ashamed that I didn't come up with that myself. Many thanks.
And 5 and 8 are part of the Fibonacci numbers: 1 1 2 3 5 8 13...
@@JohnSmith-iu8cj 1 / golden ratio. The Fibonacci sequence approaches the golden ratio.
Something that needs to be mentioned is that as you increase the physical size of capacitor (and increase the voltage rating, and use a good dialectric like X7R) the dc bias become less of an issue. This is very helpful when doing parametric searches.
For instance, suppose you need about 10uf at 5volts. A 10v 805 (Samsung CL21B106KPQNFNE) will lose about 50% of its capacitance but a 25v 1206 (CL31B106KAHNNNE) will only lose 20%. Both caps cost about the same and have good graphs on the characteristics sheet on digikey.
The recent explosion of high value small ceramic capacitors has brought this issue to the forefront. The phenomenon is due to polarization of the dielectric, and it tends to get worse as you increase capacitance and voltage rating and decrease the size of the cap. Typically as another commenter said, if you have two caps with the same capacitance and voltage rating, the smaller one will have a bigger DC bias dependency.
Part of the reason is that in order to get high capacitance and voltage rating in a small package, they switch to dielectrics like barium titanate that, while withstanding higher voltages over a smaller distance, are much easier to polarize.
A phenomenon I've noticed over 25 plus years of servicing but I never knew why until now. I now know that it's not my imagination playing tricks on me. Thanks Dave! You've out did yourself...!
I just heard about this the other day but had NO idea it had such a significant effect!
Thanks for the video.
I was going to try these type of caps in vintage radios and TVs! So pleased I watched this first.
Super job, Dave. You're right: not even a lot of "experienced" EE's know about this.
Wow
Thanks Dave....You managed to explode my head on what I thought I knew about ceramic Caps. Thanks for bringing me up to speed. Now I feel smarter.
To stop your caps misbehaving, threaten them before insertion. I find this works for me.
What do you mean? Please detail it how to do. Thanks.
Reminds me of a circuit we built for a customer that used lots of RC delays that kept failing on test because the designer had not allowed for the bias voltage of the transistors that switched on the RC circuit. We also had trouble with the dielectric aging, e.g. x7r can change by 2% after 24 hours after its been soldered (and heated above the curie point)
That's indeed very much true. Apart from the short term phenomenon Dave demonstrated in MLCCs, there is als a long term decrease of capacitance, I measured it myself during a month and the capacitance does indeed keep decreasing. When you heat up the capacitor afterwards, it gets its original capacitance back.
I read somewhere for X7R and X5R the loss is 2.5% per decade hour and for Y5V it is 7% per decade hour
Hugo Coolens Never a good idea to use x7r caps in timing circuits, but it is good fun showing the calculations to the designer on why his circuit will not meet the test limits he has specified.
Awesome video Dave ! Already faced it with one of my designs. Thanks a zillion to make it so clear.
I really liked how you superimposed the video frames in 18:23. brilliant!
I was hoping that would work in editing, and it did work nicely!
I was a product development electronics engineer for almost 20 years and never knew this. A real eye opener. When actual capacitance values matter ... pay attention!
Thanks Dave! This video really helped me with some op-amp filter design problems. Thought it was flat-out sensitivity rather than DC bias voltage. You're awesome!!!! KEEP IT UP!!!!
Check out this from Doug Ford about this issues in X7R caps in and audio filter/delay:
www.dfad.com.au/links/DFAD_PASSIVE_SURROUND.pdf
Great video, thank you very much! And very useful: the 555 timer has voltage on C changing from 1/3 Vcc to 2/3 Vcc, so it's always with that bias, and never the full capacity.
Mac Carter: Even though Voltage is relative, when you have a capacitor with a bias voltage, the dielectric has an electric field applied to it by the bias voltage, not the difference as you would expect... This electric field causes the dielectric to expand, and in capacitors the farther the plates are from each other the less capacitance you get.
That was an eye-opener, for sure. Excellent video, as always. Dom
Holy crap! 80% tolerance? I've never imagined that such caps even exist!
Nice function gen. ;-)
Truly of the best teachers, in depth knowledge and fine details being explained, beautiful lives and better world building depends on such beautiful teachers, May God bless 🙌 with more such teachers, God bless you Sir..
That is why you don't want to use ceramics in audio applications. They really add distortion. ;)
You said electrolytics do not have the issue, but they do. Probably not as much as they used to though, I seem to recall most changes happen under 1V. I no longer see cap/volt curves in the data sheets.
I remember old radios had electrolytics in their AGC circuits reverse-polarized. They got more capacitance out of it and there was something about the non-linear response.
I had a client use some MCC ceramics in a headset amp as they had those in stock instead of the ones I specified. They made great microphones ;)
They can be used, but are quite expenive for the voltage/capacitance ratings (the drawback is the piezo-effect), the only thing to keep in mind is to use a higher capacitor voltage ratio, for exemple on some of my designs I use 50V capacitors on 5V filtering and they only have a tolerance deviation of just 10% and this covers the Voltage deviation+ temperature deviation , the total maximum deviation it should be 10%(initial)+10% (temp, Volt)....also ceramics have low esr..a 10V 10uF should have a few mOhms...
I use class 1 ceramics on audio. They don't have the peizo effect.
Old video, but one of your most relevant still!
Unfortunately this video did not save me from making a split second decision for bom changes due to availability and ended up with my capacitance being reduced by 75% at my supply voltage. Needless to say, supply-filtering was not great.
BTW Murata, TDK and Tayio Yuden do show C vs DC bisa in their datasheets.
I just faced with this issue yesternday Dave :) then You know what I used? I found some piece old (from the farao's age i think) hand soldered biffy big bug cap (i will send few pieces to You, interesting!) and bang on. no changes on nothing :) the circuit is used for detect mm size metal thinks in the wood, and the oscillator is working in a wood workshop. dust, AC kicks, temperature changing, everything. And today I changed the full oscillator to tube. bang on. with modern caps, even not detect nothing :) with old caps... well almost detect a bug inside the wood just with different reaction voice. go figur! off course in this case the size of the osc is not mater:)
I have struggled for the past couple of weeks trying to make a frequency circuit that is able to go from 50hz-20khz and I have had stability problems with every circuit. The whole time I was using the wrong caps. Thanks for your help, and turning me in the right direction.
I had no idea; Great video Dave!
greate episode - I could have used that informations two week agos.
We have new component engineer at my place and I as an PCB Designer Engineer was forced to give an introduction into... well: Electronics. As I am not that deeply into components, I was the wrong pick but I guess the best option given the alternatives.
Now the fun part next on the road side: My manager wants to use altium for circuit simulation.
Well, don´t get me started with Intergrated circuits, but based on this information - all things ignored about Altium sucking at simulation - even capacitors can not be simulated with reasonable effort and/or accuracy.
Fascinating, I have been aware of some of the strange characteristics of various types of capacitor over the years but had totally overlooked this effect. I was aware of the piezo electric effects some capacitors exhibit, the self resonant effects seen in spiral wound aluminium electrolytics and some other strange memory effects.... The effect you demonstrate has serious repercussions if a coupling capacitor carrying dc bias is used to also provide LF roll-off. Thank you for uploading this.
18:30
WOW, you put a lot of effort on this.
Thanks Dave! Thumbs Up.
7:08, when you explain the data sheet, the 1+/-0.1 kHz does not mean that the capacitance of the part should be measured at 100Hz, but with a frequency between 900 and 1100 Hz, with a Voltage of 1.0 Volts.
Still, these huge variations, not just on the capacitances, but on the charging waveshape as well, are a huge surprise to me. Never expected anything like this!
Is it weird that I'm watching this video while wearing the same T-shirt that Dave is wearing in the video?
LOL!
Prevaricating parameters - Oh My!
fanboy ? :)
You forgot to add the Dave CAD logo. ;)
DaveCAD can't display the logo when disk space is too low apparently. He needs to free up some disk space.
Nice work Dave ! thanks !
Cheers Dave. Not only do they not tell you on the data sheet, they don't tell you at Uni either!!
Fantastic Video, Sir.... Its really a big trap, and before watching this I was also unaware of it....
Thanks for making such a video...
Excellent video! Didn't knew that the difference can be that big
Can be a very dramatic difference if Murphy gets you.
Fascinating indeed. That's why no education should be based solo on textbooks. The lab experience is fundamental.
There's nothing to stop this being in a textbook. For example, it's in The Art of Electronics (albeit in the X Chapters volume). The trouble is, you have to actually have to read the book... and who has time for that! (OK, I realize this comment is from 9 years ago hahaha.)
Thanks a lot for that one, Dave you are great!
nice one, radio guys learn these quirks early on, but these days speeds of even PSUs switch speeds are rising so high that almost anyone can get caught out even on small caps in mundane circtuis, never mind in RF oscillators or timing kit.
Another bugbear of mine is people not undertanding the workings of electrolytic caps. Aluminium ones, in particular, vary capacitance with applied voltage, so if undervolted compared to their working voltage rating can produce significantly lower capacitance than the marked value. I cringe when "badcaps" people say: "so I desoldered the old bad 100uF 16v and fitted a 100uF 63volt, because that'll last longer" then wonder why it fails after 2 month of huge ripple and bad hum :|
One of your best, most informative videos.
Part of the error is due to the voltage droop in the square wave due to charging current drawn. Add the output resistance of your function generator (which is also in series with your 1 k) to get a better time constant measurement. Typically, this is 50 Ohms so C=10.6 ms / 1.05kOhm gives a much more likely 10.095 uF. This is independent of the fact that ceramics' dielectric constant is electric field dependent. Also, V(t)=V_start + DV(1 - e^(-t/RC)).
Great video - reminds me of why I specialised in digital electronics back in the day :)
Another issue people miss is the fact that if a voltage is applied to a cap , an equal and opposite voltage appears on the other terminal, this is for so called dc blocking cap applications. If it it is connected to a high impedance source this an issue. The op amp or circuit can still be damaged by a dc voltage causing a negative voltage spike.
Hi Dave :-)
I think that the formula at 14:17 (with the annotation fix) is wrong.
The general (and correct) formula for the voltage across the capacitor in this circuit is :
Vc(t) = Vc(+inf) + [Vc(0+) - Vc(+inf)]*e^-(t/thau)
Where Vc(+inf) is Vfinal and Vc(0+) is Vstart. This is derived from the differential equation of the RC circuit.
This video changed my world. omg!
thanks Dave ! i was wondering why the modern 100nF caps (apparently class2 according to the dimensions) i use for decoupling have very similar decoupling effect like my old stock (15+ years) of 10nF ones ;) now i know the answer , so thanks again and keep going !
jeezes, i really am shocked! had no idea it could be so bad! thanks dave
Holy crap! I just made RC filters for my project with all caps being ceramic instead of electrolytic and this video pops up.
UA-cam starts to scare me! D:
Finally another friday fondumantal.
Keep doing this videos they are very interesting
You´re just pointed out why we do not use ceramic caps in any audio circuits path.
As the insulator between the foils witch forms the cap is changing its values over bias voltage....
Interacting to get this on more people's feed again
I think the term for this is "derating" - so the graph Dave showed of change in capacitance against voltage is a derating curve. Apologies if someone else already said this!
Good to know. Thanks Dave!
Didnt follow your fund. Fridays for some days. These vidoes are always my bored time killers !
Interesting. Can you do a video on how one would work around this?
Great video, I knew there were some variances in these caps by looking at datasheets but it would be interesting to know more about the physics of why they change so drastically under these conditions. The materials used under DC bias conditions and various scenarios which lead to this variance would be a good video to get more in depth with the processes going on.
It is basically the electric dual of magnetic saturation. To make high-value capacitors physically small you use dielectrics with high relative permittivity (e.g. Barium Titanate). There is no free lunch though, the extra polarisation comes from physical reconfiguration of the charges in the dielectric material, and there is a limit to how much you can polarise a dielectric before it can help you no more (saturation - a state as ordered and contributing to the applied field as it can be in its current phase).
vk2zay Thanks for the informative reply.
Great demonstration! This can really bite you if you are using caps to absorb transient voltages. We use ceramics right at connectors for EMI and ESD protection. I have been bit by this before in the past because i didnt compensate for this drop in capacitance and the caps kept exploding. Frown!
Was having this issue in the grid feed inverter I'm working on. I put some RC snubbers on the MOSFETs and could only find ceramics at hand, and it didn't improve the ringing as calculated (did reduce it, but it was still there).
After ordering in some polypropylene film caps (Vishay MKP385), now no ringing! I suspected that it was this effect along with better temperature characteristics of poly caps perhaps.
Just Wow.... Thank you so much for this Dave! What would you recommend if you do need to use a cap in a timing circuit? A Class 1 NPO?
Absolutely. Or a poly cap.
NP0 only.
Very interresting, they seem to change more than varicap diode.
There are probably some interesting applications for deliberately exploiting this property and using them as varicaps :-)
Judd Niemann
Probably, but likely just as crap at that as they are at being capacitors :->
Variable naming in the formula at 14:17 isn't right. if Vstart = 0, then you the multiply the entire expression by 0, and the formula says the voltage will always be 0 which clearly isn't right. For the step input, V0 isn't zero but the step voltage i.e.5 volts in this example; plot y = 5(1 - e^(-t/RC)) and that'll be clear.
My guess for the DC offset is that it should be V = Voffset + Vstep(1 - e^(-t/RC)), i.e. 5 + 1(1 - e^(-t/1000C)) in this case...?
Oops, I forgot to add the subtraction from the final value. Vo=Vfinal - Vstart
Fixed in annotation.
7:15 you say "0.1 kHz, 100 Hz", but it is 1+-0.1 kHz which is 900-1100 Hz.
Small, cheap, stable - pick any two.
The only truly linear dielectric is vacuum, and even that will eventually screw you with tunnelling at close spacing or vacuum breakdown and pair production at more extreme conditions.
Thanks for that Dave, but I wonder, could you do a video about how to choose the thickness of a wire? I am having some issues with it. (according to an online calculator it should'ave worked, but the wire ended melting the insulation, shorting over and destroying my brand new digital temp sensor...)
Great info Dave!!!
Very informative, thanks. Did you, by chance, decide to address this topic based on my recent comment on one of your older videos about caps where I warned about this problem?
No, I don't recall that. It's been on my to-do list for a very long time.
EEVblog Never mind. The minds of people of a feather seem to be surprisingly aligned sometimes. Not that I'll ever be in the same league as you. Amazing job.
Dave can you give some of your thoughts on the red pitaya instrument, it looks very interesting and your opinion would be very valuable
learned a new thing. seems like that could be a feature? Like a VCO?
Awesome episode
thnx DAVE!!!!!
I think you misread the table at 7:13. It should have been the second row. Also, the cap should have been tested at 1khz+-0.1 So between 900 and 1100hz, whether it was the first or second row. The voltage should have been 0.5v+-0.1. I think so anyway! Sorry to be so nit-picky. But it's what I'm like.
6:36 "102 is 10 microfarads". Actually, 102 is 1000pF which is 1nF. For 10 microfarads the code is 106, as seen in the part number at 5:28.
what is the best capacitor type for high voltage?
thanks for telling me that "there is some weird...you know... physics going on here" in the y5v caps :D. Great Video
That is such a great tutorial
Reminds me of how the gate charge on a MOSFET behaves...
The gate appears as a capacitance, but that increases with the voltage,
so whilst the gate may appear as ~1nF at a 0-5V signal, a 0-7V signal may see a ~3nF capacitance;
this is usually written as "gate charge" measured in coulombs, and is typically given at 5V. -some- datasheets give an additional level, maybe 10V...
The increase in gate charge is pretty drastic, and not many datasheets have it marked outside the standard 5V level, and i'm yet to find a graph
that illustrates the gate charge Vs gate voltage characteristic.
And as dave said, don't even mention the effects of frequency... That's just a whole dimension of changes on it's own :/
That was really informative. I wonder if these effects are modeled in simulation packages, like LTspice.
Hi Dave, warm regards from cold Russia! I tried to verify your info with my RLC meter by splitting voltage on 2 caps and it is true, the capacitance drops tremendously. If you have 5 minutes you can look at my video report named "Ceramic Capacitance" on my channel (can't add video reply though). Thank you for sharing knowledge with us!
Dave, I think you have a little typo in the vid description "If your 10uF capacitor really 10uF in your circuit?" but I think you meant Is. ;)
Btw, great video as always! Keep up the great work :)
You can put that 6% down to the tolerance on the resistor too, no? I didn't see you measuring the 1k resistor's actual value, which definitely plays into the calculation of tau. Not that it makes a significant difference to the point you're making, of course.
As much as this is super interesting. This methodology is flawed. You cannot really use RC constant method, if the charging is not in the shape of 1-e^(-tau/t). What you can do instead, is take a log(1 - V/Vfinal) and fit a line a*t+b, and from average 'a' find out capacitance over range, or log(1 - V/Vfinal)*t, and see how capacitance "changes" as it MLCC is being charged. (i.e. it is operating in higher voltage region). Similar for discharge. In fact this will show variations in "capacitance" in single shot without need to change step response in signal generator. I am certain these effects can be pretty accurately modeles as parasitic inductances and capacitances, and they relate directly to physical reasons why we see these shapes on scopes.
zomg, my E: drive filled up while watching this. Lol, you gave me quite the scare there... full screen at 22:07
Dave, check up this video TDK Tech Tube: Measuring Capacitance accurately in an MLCC
Seems to me manufacturers should give both the max voltage range AND the bias voltage where the cap drifts out of spec by 10 or 20%.
Second link (avx) now a deadlink on avx site.
The next avx link has similar problem. Wawawah
That helps explain why my 3 means used to measure ceramic caps vary so much.
Thanks!
Great video.
And this is why sensible people only use multilayer ceramics for decoupling & bypass purposes. ;)
I'm thinking those low-grade ceramic caps could double as AC-no-Vf varicaps.
in the 5V biased case, the voltage applied across the cap is the same 6V, but according to murata's QA here www.murata.com/support/faqs/products/capacitor/mlcc/char/0005 and a lot other threads they treat 'DC bias voltage' as voltage across capacitor, and capacitance decreases when voltage increases, well in Dave's case the voltage across the cap stays the same, so apparently they are not the same thing, or are they?
You didn't test the capacitor at voltages and frequency according the manufacturer's specifications. The data sheet at 7:15 specifies that test RMS voltage should be between 0.4 and 0.6 V RMS (line 2 column 2 for a 6.3 V cap). You tested the caps at 5 V square wave with duty cycle of 20% which corresponds to an RMS value of 2.24 V. Four times the recommended voltage. According to column 1 line 2 it should also have been tested between 900 and 1100 Hz. You tested it at 5 Hz? Now that's a greater challenge if you want to see the entire charge cycle on the scope but would be possible at 1 kHz with a 10 ohm resistor given a 10 microF cap.
Compare a 0v to +6v pulse on a capacitor to a -3v to +3v pulse on a capacitor.
Big thumbs up!
Dave,
Was this a problem in old ceramic caps used with tube circuits?
OldTech
11:00 You are very close, did'n you forget the gen impedance? So Indstead of 1k the R should be 1.05k.
Excellent Dave! Thank you!!
And a bit of quibbling... in the case where you had a 0 to 10V, 20% duty cycle waveform, you do have a bias, right? because the average value is, what? 2 V ? (Don't know the exact value... it's been too many years...)
No bias on 0-10V
EEVblog
No "external bias" per se... but the average voltage is non-zero and so isn't that an effective DC voltage?
EEVblog Your 0-10V signal is biased. It has a substantial DC component. To be unbiased you would need to swing from negative to positive in a way that eliminates the DC component.
Many electronics engineers are in total denial about the weird behaviour of capacitors, and accuse you of being an audiophile when you try to explain these effects to them. This is one of the very few areas where the audiophiles actually have it right. Capacitors are seriously funky, and it really matters which ones you use. Those DC related effects you are seeing are non-linear.
Another funky polarisation effect with capacitances happens with Kapton flexible PCBs. The strength of the effect varies with the exact variety of Kapton, but I think it is always there. Measure the capacitance of some traces on a flexible PCB. Touch the PCB for a moment and the capacitance goes up. Let go and the capacitance falls to its original value.Now touch and keep touching the PCB for a couple of minutes, and then let go. The capacitance falls as you let go, but to a value distinctly higher than its original level. Over the next minute or so the capacitance gradually falls to its original level. This is not due to your finger warming the Kapton. You can warm the Kapton with hot air to around body temperature and you won't see the kind of capacitance change a sustained touch causes. The Kapton flexible PCBs which Apple use for cap-touch pads in the ipods only show this effect to a minor extent. I think they carefully selected the type of Kapton they use for this very characteristic. I have seen other products with Kapton cap-touch sensors where the effect is so strong that the adaptive capacitance tracking algorithms are only just about able to cope with the change.
Interesting video ! Thanks :)
Those tricky little bastards. I would never guess the cause if something like this would happen in real life circuit. On the other hand, I wonder if those caps could be used like varicap diodes.
What is the best type of capacitor to use for high DC bias then?
Was that test voltage meant to be 0.5V? The cap voltage is 6.3V so I would have thought row 2 would apply.
So, would this be the basis for very wide range voltage controlled oscillators?
Ok, so the capacitance changes with offset voltage but what is really changing? The physical dimensions and construction of the capacitor plates which calculate to 10 uf are not changing, yet the capacitance is. How can this be? Is the actual internal capacitor plates physically expanding/ shrinking to cause this effect?